Herpetological Conservation and Biology 7(2): 115–131. Submitted: 15 March 2012; Accepted: 17 July 2012; Publication: 10 September 2012.
ADVANTAGES OF LONG-TERM, MULTI-SCALE MONITORING: ASSESSING THE CURRENT STATUS OF THE YOSEMITE TOAD (ANAXYRUS [BUFO] CANORUS) IN THE SIERRA NEVADA, CALIFORNIA, USA
1 CATHY BROWN , KATIE KIEHL, AND LUCAS WILKINSON
USDA Forest Service, Stanislaus National Forest, 19777 Greenley Road, Sonora, California 95370, USA 1Corresponding author, email: [email protected]
Abstract.—A comprehensive bioregional evaluation of the current status of at-risk species is critical for effective management and conservation. As part of a long-term, multi-scale amphibian monitoring program, we evaluated the current status of the Yosemite Toad (Anaxyrus [Bufo] canorus) on national forest lands across the species’ range in the Sierra Nevada, California, USA. We conducted monitoring at extensive (rangewide) and intensive (individual population) scales with small watersheds (2–4 km2), individual meadows, and breeding areas within meadows as sample units. Yosemite Toad breeding was found in an estimated proportion of 0.84 ± 0.03 (SE) of recently occupied watersheds (locality data from 1990–2001), and 0.13 ± 0.04 of historically occupied watersheds (locality data prior to 1990). Rangewide, breeding was found in an estimated proportion of 0.22 ± 0.01 of watersheds. We quantified demographic parameters (abundance and survival of breeding males, abundance of egg masses, successful metamorphosis) in six meadows in two watersheds. Abundances were small, with the largest populations having only 16–21 and 18–19 breeding males each year. Annual survival rates of males by meadow ranged from 0.49–0.72. Numbers of egg masses per year were low, and the proportion of breeding areas with successful metamorphosis ranged from 0.14–0.73. Using the multi- scale data, we examined Yosemite Toad spatial and temporal occupancy patterns. Yosemite Toads tended to breed in one or two sites (lakes, meadows, stream reaches) per watershed every year, and occasionally in other sites. This assessment of the status of the Yosemite Toad will inform management decisions at both bioregional and population scales and the development of conservation priorities.
Key Words.—amphibian demography; Anaxyrus canorus; bioregional monitoring; multi-scale monitoring; Sierra Nevada; Yosemite Toad
INTRODUCTION al. 1991; Hecnar and M'Closkey 1997; Meyer et al. 1998; Skelly et al. 2003; Trenham et al. 2003). Amphibians are imperiled worldwide (Beebee and Logistically, the types of data that are feasible to collect Griffiths 2005; Wake and Vredenburg 2008) and over at bioregional and local scales differ (see Beebee and 40% of all amphibian species have decreasing Griffiths 2005). Detailed abundance and demographic population sizes (Stuart et al. 2004). As efforts to data provide the most insight into population dynamics, identify and monitor at-risk populations have increased, but are usually too expensive and logistically impractical the role of spatial and temporal scale in providing robust to collect long-term and over large geographic areas. As assessments of status and understanding the population a result, researchers increasingly use occupancy data as a dynamics of amphibians has come to the forefront more cost-effective alternative (Trenham et al. 2003; (Hecnar and M'Closkey 1997; Meyer et al. 1998; Skelly Bailey et al. 2004; Corn et al. 2005; Gould et al. 2012). et al. 2003). Many population studies have been However, occupancy data alone provide no information conducted at local scales such as individual lakes or on changes in abundance, which may be a first sign of ponds and often over short time periods (Briggs and impending distributional declines (Muths et al. 2003; Storm 1970; Pechmann et al. 1991; Richter and Siegel Weir et al. 2009; Collen et al. 2011), or key vital rates 2002). However, bioregional assessments, generally and life-stage specific population dynamics (Berven conducted over large areas often defined by ecological 1990; Semlitsch et al. 1996; Biek et al. 2002; Scherer et characteristics, are needed to fully evaluate a species al. 2005). One solution is to collect occurrence data status (Stuart et al. 2004; Gompper and Hackett 2005; across bioregional scales to evaluate rangewide Bonardi et al. 2011; Cameron et al. 2011; Gould et al. persistence and more detailed demographic data at a 2012), and monitoring over broader spatial and temporal smaller subset of locations to provide insights on local scales may be more effective at separating general population dynamics such as abundance, survival, and population trends from local fluctuations (Pechmann et recruitment (e.g., Corn et al. 1997).
Copyright © 2012. Cathy Brown. All Rights Reserved. 115
Brown et al.—Current Status of the Yosemite Toad.
The Yosemite Toad (Anaxyrus [Bufo] canorus) is a lentic water) within sites. In this paper, we present a species well suited for a multi-scaled monitoring baseline assessment of bioregional population status for approach. The Yosemite Toad is endemic to the Sierra the Yosemite Toad. We also show how data collected Nevada, California, where it occurs predominantly on for status and trend can be used to address other public lands (almost 99% of its range), much of which is questions useful for management by presenting insights in wilderness areas. Declines in the Yosemite Toad have on Yosemite Toad spatial and temporal occupancy been documented (Kagarise Sherman and Morton 1993; patterns. Jennings and Hayes 1994; Drost and Fellers 1996), and the U.S. Fish and Wildlife Service has determined that MATERIALS AND METHODS listing the species as endangered under the Endangered Species Act is warranted but precluded by higher Study region.—The study region for the extensive priority actions (U.S. Fish and Wildlife Service 2002). component is the portion of the historical range of the However, a comprehensive assessment of status across Yosemite Toad that occurs on national forest lands the species' range has not previously occurred and within the Sierra Nevada, California, USA (Fig. 1). This potential causes of decline are not well understood. area comprises about 10,500 km2 extending from Alpine There are multiple challenges to monitoring the County south to Fresno County and contains survey Yosemite Toad. If surveys are timed correctly, all life watersheds at elevations from 1,859 to 3,703 m. Aquatic stages are relatively easy to detect; however, the suitable habitats within the study region are chiefly fed by survey periods for each life stage are relatively short and snowmelt, and include lakes, ponds, wet meadows, and are not concurrent. Breeding commonly occurs in streams, which vary greatly in size, vegetation, shallow, warm water areas that typically dry over the hydrology, and topography. A variety of management summer and include wet meadows, shallow ponds, activities occur within the study region although much is shallow flooded, grassy areas adjacent to lakes, and slow in designated wilderness where management activities flowing streams (Karlstrom 1962). Yosemite Toads are are more limited and access is only by foot or pack explosive breeders where adults arrive at snowmelt, lay animal. Livestock grazing occurs in many Sierran eggs over a short period of time, and then leave the meadows and legacy effects exist even in those areas breeding areas (Kagarise Sherman 1980). Because they where grazing no longer occurs (Menke et al. 1996). are not commonly found after breeding, surveys for Fish stocking has occurred since the late 1800's and is adults must occur at snowmelt over a 1–2 week period, still widespread (Bahls 1992; Knapp 1996), and when most locations are difficult to access. Eggs hatch recreational activities are increasing in high elevation into larvae in about 4–15 days (Karlstrom 1962; wilderness areas (USDA Forest Service 2001). Other Kagarise Sherman 1980), tadpoles often metamorphose activities such as vegetation, fire and fuels, and road by early to mid August (Karlstrom 1962; pers. obs.), and management also may affect the species at lower metamorphs disperse from immediate breeding areas elevations. (pers. obs.). The rapid development, desiccation of The two intensive watersheds included Bull Creek ephemeral breeding areas, and dispersal behavior result (4.86 km2), located in Fresno County in the southern in a short survey window for each of these stages, which, Sierra Nevada, and Highland Lakes (2.9 km2), located in combined with the remote locations of much of the Alpine County in the central Sierra Nevada (Fig. 1). species’ range, makes it logistically difficult to collect Elevations in Bull Creek range from 2,134 m to 2,489 m demographic information on all life stages over and vegetation is predominantly forested with small bioregional scales. meadow and riparian openings (0.33 ha to 5.1 ha). To provide a comprehensive bioregional assessment of Management activities occurring in the watershed the status of the Yosemite Toad that addressed these include livestock grazing, recreation, timber harvest, and monitoring challenges, the USDA Forest Service ongoing ecological research. Highland Lakes, situated implemented a long-term multi-scale monitoring in a saddle surrounded by high peaks, is chiefly open program on national forest lands in the Sierra Nevada. with large areas of alpine meadow containing numerous In an extensive component, we estimated the proportion small pools and areas of flooded vegetation. Elevations of occupied watersheds at the scale of the range of the of suitable Yosemite Toad habitat range between 2,536 Yosemite Toad on national forest lands using a spatially m and 2,621 m. This watershed contains a popular balanced, unequal probability survey design. In an campground and draws large numbers of visitors for intensive component, we estimated abundance and both summer and winter recreational activities. survival of adult breeding males, number of egg masses, Livestock grazing also occurs in the watershed. and successful metamorphosis in six meadows in two watersheds. Between the two components, we Design and field methods extensive component.— conducted surveys at three scales, watershed, site (lake, Probability-based survey designs such as the one used in meadow, stream reach), and breeding area (contiguous this program require a precisely defined target
116
Herpetological Conservation and Biology
FIGURE 1. Study region and distribution of known occupied watersheds for the Yosemite Toad (Anaryxus canorus) in the Sierra Nevada, California, USA. The monitoring area is national forest lands within the range of the Yosemite Toad. Bull Creek and Highland Lakes are the two Yosemite Toad intensive watersheds
population (i.e., the set of units defining the geographic not feasible to create one (i.e., delineate all 2–4 km2 scope of inference) and a sample frame (commonly a watersheds within the monitoring area), we used the geographic information system [GIS] representation of Forest Inventory and Assessment (FIA) systematic the target population) from which the sample is selected random hexagonal grid (Brand et al. 2000) densified to (Olsen et al. 1999). We defined our target population as hexagons of 2.6 km2. all watersheds, delineated to 2–4 km2 in size, with We compiled all available Yosemite Toad locality data available aquatic habitat on national forest lands within from state and federal agency records, academic the range of the Yosemite Toad. Because there was no researchers, and from literature and museum sources. existing sample frame (e.g., GIS coverage) and it was We assigned each of the 10,077 hexagons (and
117
Brown et al.—Current Status of the Yosemite Toad. corresponding grid point centers) in the sample frame to Design and field methods intensive component.—We one of three mutually exclusive temporal categories subjectively selected two intensive watersheds using the based on the compiled locality data: (1) Recent, toads following criteria: (1) contained a variety of aquatic documented within 10 years prior to the start of the habitat types; (2) contained at least three breeding lakes, sampling (defined to be after 1990); (2) Historical, toads meadows, or stream reaches separated by more than 200 documented only prior to 1990; and (3) Unknown, no m; (3) supported a relatively large number of adult toads; locality data. A spatially distributed random sample was and (4) had easy access at snowmelt and during the selected from the sample frame using a generalized summer. From 2006–2009, we visited three meadows in random tessellation stratified survey design (GRTS) for each watershed at snowmelt to conduct adult capture- a finite resource with unequal probability of selection mark-recapture (CMR) and egg mass surveys (Bull based on the three temporal categories (Stevens and Creek meadows: 520M15, 520M16, 520M20; Highland Olsen 1999; Stevens and Olsen 2004). We selected a Lakes meadows: 37188, 37213, 37165). The ecology of larger proportion of the sampled hexagons from areas the Yosemite Toad is ideal for using Pollock's robust most likely to have Yosemite Toads (Recent). We design (Pollock 1982; Kendall et al. 1997) to estimate delineated watersheds, 2–4 km2, around the selected grid abundances of breeding males during their spring point centers using the Hydrologic Unit Code rule set chorus. The robust design is two-tiered with primary (U.S. Department of Agriculture 1997). sampling periods that occur over longer intervals that Temporally, the study design uses an augmented allow population gains (recruitment/immigration) and serially alternating panel design (Urquhart et al. 1998; losses (mortality/emigration), and secondary sampling Urquhart and Kincaid 1999). The initial design specified occasions within each primary period where the that 20% of the Recent watersheds (n = 16) were to be population is assumed to be closed. We treated each visited every year (referred to as the annual watersheds), annual breeding period as a primary sampling period. and the remaining watersheds once every five years on a Because males congregate in breeding choruses for the rotating schedule. Due to funding limitations, the full short duration of the breeding period, we treated multiple sample of watersheds took eight years to complete samples over consecutive days during each primary (2002–2009). Based on preliminary analyses of the first period as secondary occasions from a population that is five years of data, we determined that completing the relatively closed. The robust design is not as suitable for full sample, thereby increasing the precision of the estimating female abundances. Female toads remain in estimates, was more important than beginning the breeding areas for only a few days leaving shortly after second cycle. amplexus, thus violating the assumption of population We determined watershed occupancy by surveys of all closure among the secondary occasions. Still, we lentic (lakes, wet meadows) and a sample of lotic habitat marked females when possible. (perennial and ephemeral streams) within each We conducted CMR surveys as soon as breeding watershed. We identified lakes and meadows using choruses formed at snowmelt and surveys continued for aerial photographs that we verified in the field. We at least five consecutive days each year. The timing of randomly selected three 300-m perennial stream reaches snowmelt and the subsequent breeding varied among per watershed, one from each of three gradient classes years ranging from late April to mid May in Bull Creek (low < 2%, medium 2–8%, and high > 8%), and two and from late May to late June at Highland Lakes. We 100-m ephemeral stream reaches located in meadows surveyed each intensive meadow for at least six hours from a USDA Forest Service Sierra-wide GIS coverage each day. At each meadow, crews rotated among the (Region 5 Remote Sensing Lab in 2001, available from breeding choruses, visually scanning the breeding areas, Cathy Brown). We refer to the individual lakes, actively searching in burrows and under ground cover, meadows, and selected stream reaches in each extensive and using cues from calling males to locate toads. All watershed as sites. unmarked toads > 40 mm snout-urostyle length (SUL) From June-September 2002–2009, we used Visual were PIT-tagged (passive integrated transponder; Martin Encounter Surveys (VES) to determine Yosemite Toad 2008) using AVID MUSICC MicrochipsTM (Avid occupancy in each extensive watershed (Crump and Identification Systems, Inc., Norco, California, USA) Scott 1994; Olson et al. 1997). We surveyed all and released. Upon initial capture each year, we wadeable water at each site (lake, meadow, and selected recorded the snout-urostyle length, weight, and sex of stream reach) and we made an effort to survey each toad. We identified marked animals by their watersheds early in the season before sites dried up to unique tag numbers at each subsequent recapture, and maximize detection of the tadpole stage. Although our we used these data to create capture histories. primary goal was to detect tadpoles, we recorded data on We conducted egg mass counts at the end of the five all life stages observed. day adult CMR period. Conducting counts of Yosemite Toad egg masses can be challenging because females lay eggs in long strings and sometimes split masses over two
118
Herpetological Conservation and Biology or more locations or lay them communally. Also, eggs moving through an area. Therefore, our primary goal may hatch prior to the completion of the five day period was to estimate the proportion of watersheds occupied or be deposited after counts are conducted. To address ( ) by breeding (eggs, tadpoles, or metamorphs). Too these challenges, we used several methods to determine few adults were found to model them separately, so to the most efficient approach. We used the more intensive account for all areas occupied by the species, we also dependent double-observer method (Grant et al. 2005) in estimated the proportion of watersheds occupied by any 2006 and independent double-observer method (Nichols life stage. For both occupancy states (breeding, any life et al. 2000) in 2007 and 2008. After employing and stage), we estimated the proportion of occupied testing these multiple observer methods, individual egg watersheds for the entire range of the species mass detection probabilities were very high (> 0.96) and (rangewide, (.) ) and relative to the temporal period consistent among observers. Accordingly, in 2009 we when toads were last observed (Recent, Historical, simply used a tally of unique eggs masses (i.e., Unknown, (temporal category) ). We used a maximum conditional probability of detection = 1) from a single likelihood approach that incorporates the probability of observer. This was simpler and more efficient, and in detecting the species (p) and the watershed weights our judgment, an accurate estimate of the number of (described below; also MacKenzie et al. 2006). Methods Yosemite Toad masses laid each year. Although the that do not account for p typically result in an methods differed among years, the estimates of egg mass underestimate of occupancy unless the probability of numbers were similar and reasonable given the numbers detection is near one (MacKenzie et al. 2006). We of adults we observed. Thus, in our judgment, the three assumed that the occurrence of toads was closed during methods were comparable. In addition, based on our the eight-year monitoring period (i.e., occupancy state in observations during the five day period and our each watershed did not change) and used the annual subsequent surveys of these meadows for the extensive watershed data to estimate the probability of detection. component, it is unlikely that we missed large numbers This approach differs from most studies where visits of egg masses. occur over shorter time periods, typically within a In 2006, we visited each intensive meadow every two season. If the population closure assumption is violated, weeks (Bull Creek 29 June - 25 August; Highland Lakes the probability of occupancy may be overestimated. We 12 July - 7 September) for a total of six visits. At each recognize that eight years may be a long time to assume breeding area in each intensive meadow, two observers that watershed occupancy is constant. Our approach independently counted tadpoles and metamorphs. may be reasonable given the likely slow and persistent Additional surveys were conducted by one observer at effects resulting from many management activities, but Bull Creek once per week (10 visits, 17 June - 19 be more questionable for threats such as an epidemic August) in 2009 and at Highland Lakes once every two pathogen that causes widespread and rapid declines. We weeks (five visits, 15 June - 27 August) in 2007 and address this assumption and its implications in our twice (14 July, 29 July) in 2008. These data were used discussion. to document successful metamorphosis for each We modeled p as a function of two covariates, survey breeding area. date (measured by the number of days from January 1) and snow pack of the preceding winter. Snowpack was Analysis extensive component.—We based the quantified as the percentage of normal snowpack for the analyses of occurrence data on an unequal probability Sierra Nevada measured on 1 April each year (California survey design where the weight for each watershed Department of Water Resources, California Data evaluated is the inverse of the probability of selection for Exchange Center. Available from http://cdec.water.ca. that watershed (Diaz-Ramos et al. 1996). The gov/ [Accessed 17 September 2009]). The Sierra contribution of each sample watershed to the final Nevada receives most of its precipitation between estimate is proportional to the watershed weight. For December and April (Howat and Tulaczyk 2005). Date example, sample watersheds with higher probability of of survey and seasonal water are important factors for selection (e.g., Recent watersheds) have lower weights detecting many amphibian species (Schmidt 2005; Weir and thus contribute less to the final estimate. We et al. 2005; Lacan et al. 2008). Yosemite Toads breed in conducted analyses using the R software (R ephemeral habitats and metamorphose within one Development Core Team, Vienna, Austria), an R summer. Thus, we predicted that p should be highest in contributed library (spsurvey: Spatial Survey Design and early season surveys and in years with higher snow pack Analysis, Kincaid and Olsen 2009) developed for the (i.e., more available water). Habitats dry up sooner in statistical analysis of probability survey data, and SAS® drought years, so we tested for a survey date and snow software (SAS Institute Inc., Cary, North Carolina, pack interaction. We developed ten a priori models USA). where ψ = probability of occupancy and p = probability The presence of tadpoles indicates the existence of a of detection (Table 1). reproducing population rather than, for example, an adult We used a generalized version of the maximum
119
Brown et al.—Current Status of the Yosemite Toad.
TABLE 1. Model selection for occupancy estimates for the Yosemite Toad (Anaryxus canorus) for breeding and any stage, summarized rangewide and by temporal category. K is the number of parameters in the model. AICc is the Akaike Information Criterion adjusted for small sample sizes. wi is the Akaike weight for comparing models. ψ is the probability of occupancy and p is the probability of detection. Models comprising ≥ 95% of the weight are in bold.
Breeding Any Stage
Model K AICc ΔAICc wi Model K AICc ΔAICc wi Rangewide, ψ (.) p(Survey Date * Snow Pack) 5 1851.7 0.0 0.996 p(Survey Date * Snow Pack) 5 2059.9 0.0 1.000 p(Snow Pack) 3 1862.8 11.1 0.004 p(Survey Date + Snow Pack) 4 2232.0 172.2 0.000 p(Survey Date + Snow Pack) 4 1885.2 33.5 0.000 p(Snow Pack) 3 2276.0 216.2 0.000 p(.) 2 2027.4 175.7 0.000 p(Survey Date) 3 2368.6 308.8 0.000 p(Survey Date) 3 2029.3 177.6 0.000 p(.) 2 2372.0 312.2 0.000 Temporal Category, ψ (Temporal Category) p(Survey Date * Snow Pack) 7 1459.8 0.0 1.000 p(Survey Date + Snow Pack) 6 1872.7 0.0 1.000 p(Snow Pack) 5 1487.8 28.0 0.000 p(Snow Pack) 5 1909.8 37.2 0.000 p(Survey Date + Snow Pack) 6 1523.4 63.6 0.000 p(Survey Date * Snow Pack) 7 1913.8 41.2 0.000 p(.) 4 1534.1 74.3 0.000 p(.) 4 1955.4 82.7 0.000 p(Survey Date) 5 1538.6 78.8 0.000 p(Survey Date) 5 1957.6 85.0 0.000 likelihood procedure developed by MacKenzie et al. (2002) appropriate for our more complex unequal 1 probability survey design. The following notation is used: S, the number of space-time subunits in the study; where S subunits are present and each watershed has a n the number of watersheds in the sample in subunit s; s, weight, w and V, number of distinct seasonal sampling occasions; , si, si 0 if ′ 0 probability that a species is present in watershed i within ′ 1 if 0 subunit s; psiv, probability of detecting a species in watershed i within subunit s on visit v, given the species and both si and psiv may depend on the covariates. The term distinguishes the watersheds where a is present. Then let , ,…, where species is detected on at least one visit and those sites 1 if species is detected in watershed i in where a species is not detected on any visits. This is the subunit s on visit v same likelihood equation as in MacKenzie et al. (2002) when all weights are equal, as is the case for a simple d = 0 if species is not detected in watershed siv i in subunit s on visit v random sample. We maximize the log likelihood using NA if no visit was made in watershed i in the function “optim” in R (R Development Core Team, subunit s on visit v Vienna, Austria). Variance estimates for and are obtained from the Fisher information matrix derived and s=1,…,S, i=1,…,ns and v=1,…,V. Let from the Hessian matrix evaluated at the and , ,…, be K watershed-specific covariates estimates. Confidence intervals are constructed as the for watershed i within subunit s and estimate ± the square root of these variance estimates multiplied by the 1 ⁄ 2 th percentile of the standard normal distribution. We evaluated the relative support for each model using Akaike Information Criterion for small sample sizes (AICc) and Akaike weights (wi) (Burnham and Anderson 2002). We dropped models be M watershed-specific and visit-specific covariates for that could not be fit to the data from the model set and watershed i within subunit s. The likelihood equation we used the remaining models to compute model- for a stratified unequal probability survey design, such as averaged estimates (Burnham and Anderson 2002; the one used in this program, is MacKenzie et al. 2006).